ATINER's Conference Paper Series IND2016-1974ATINER CONFERENCE PAPER SERIES No: IND2016-1974 5...
Transcript of ATINER's Conference Paper Series IND2016-1974ATINER CONFERENCE PAPER SERIES No: IND2016-1974 5...
ATINER CONFERENCE PAPER SERIES No: LNG2014-1176
1
Athens Institute for Education and Research
ATINER
ATINER's Conference Paper Series
IND2016-1974
Aslan Deniz Karaoglan
Assistant Professor
Balikesir University
Turkey
Beyazit Ocaktan
Assistant Professor
Balikesir University
Turkey
Abdullah Cicibas
M.Sc. Student
BEST Transformers
Turkey
Flow Time and Due Date Estimation for Customer
Orders According to Design Criteria: A Case Study
of BEST Transformers Company
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
2
An Introduction to
ATINER's Conference Paper Series
ATINER started to publish this conference papers series in 2012. It includes only the
papers submitted for publication after they were presented at one of the conferences
organized by our Institute every year. This paper has been peer reviewed by at least two
academic members of ATINER.
Dr. Gregory T. Papanikos
President
Athens Institute for Education and Research
This paper should be cited as follows:
Karaoglan, A. D., Ocaktan, B. and Cicibas, A. (2016). "Flow Time and Due
Date Estimation for Customer Orders According to Design Criteria: A Case
Study of BEST Transformers Company", Athens: ATINER'S Conference
Paper Series, No: IND2016-1974.
Athens Institute for Education and Research
8 Valaoritou Street, Kolonaki, 10671 Athens, Greece
Tel: + 30 210 3634210 Fax: + 30 210 3634209 Email: [email protected] URL:
www.atiner.gr
URL Conference Papers Series: www.atiner.gr/papers.htm
Printed in Athens, Greece by the Athens Institute for Education and Research. All rights
reserved. Reproduction is allowed for non-commercial purposes if the source is fully
acknowledged.
ISSN: 2241-2891
13/09/2016
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
3
Flow Time and Due Date Estimation for Customer Orders
According to Design Criteria: A Case Study of BEST
Transformers Company
Aslan Deniz Karaoglan
Beyazit Ocaktan
Abdullah Cicibas
Abstract
In the electro-mechanics industry, the manufacturing is performed by project
type production and most of the sales are performed by tender offers. Because
of project based manufacturing, the most of the transformers are produced for
the first time and the processing times for these orders are unknown. In this
case it is important to estimate the cost of the customer order before the
production at a bidding stage for accurate price offer and win the tender. In this
labor-intensive sector; accurate estimation of the labor cost has great
importance to give a realistic price offer. Therefore, it is important to estimate
the flow time and due date with a low variance. In this study, the core
production process of a transformer company is simulated by using Arena
software, and the technical specifications indicated in the customer orders are
used as simulation inputs for the transformers to be produced for the first time.
Keywords: Due date, Flow shop, Flow time, Simulation, Transformer.
Acknowledgments: This research is supported by Republic of Turkey
Ministry of Science, Industry and Technology [grant number: SANTEZ-
0682.STZ.2014]. The authors would gratefully like to thank BEST
Transformer Research & Development Department for its valuable supports
lead to reveal this paper
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
4
Introduction
In the electromagnetics sector transformers are produced by a labor
intensive project type production system. Except from a few similar
transformers produced frequently; there are unlimited types of different
transformer orders which vary according to the customer specifications those
will be produced the first time. This variation causes different processing
times, a flow time (FT) and due date (DD) for each type of order which has to
be predicted before starting the production.
There are so many studies presented in the literature that estimate the
processing time, FT, and DD by using simulation models, probability
distributions and other statistical techniques and presented remarkable results.
Ekren and Ornek (2008) presented a simulation based experimental design to
analyze factors affecting the production flow time. Baykasoglu and Gokcen
(2009) proposed a new approach that is based on a genetic programming
technique which is known as a gene expression programming (GEP), and
compared the performance of the proposed due date assignment model with
several previously proposed conventional due date assignment models. For this
purpose, simulation models are developed and comparisons of the due date
assignment models are made. Slomp et al. (2011) studied on estimation of flow
times of jobs to be produced in a flexible manufacturing cell (FMC) consisting
of a number of identical machines by the aid of simulation. Joseph and
Sridharan (2011) investigated the effects of dynamic due-date assignment
models (DDDAMs), routing the flexibility levels (RFLs), sequencing
flexibility levels (SFLs) and part sequencing rules (PSRs) on the performance
of a flexible manufacturing system (FMS) for the situation wherein part types
to be produced in the system arrive continuously in a random manner. They
used a discrete-event simulation model of the FMS as a test-bed for
experimentation. Vinod and Sridharan (2011) also presented the salient aspects
of a simulation study conducted to investigate the interaction between due-date
assignment methods and scheduling rules in a typical dynamic job shop
production system. Simulation experiments are carried out for the different
scenarios that arise out of the combination of due-date assignment methods and
scheduling rules. They found that dynamic due-date assignment methods
provide better performance and they developed regression-based
metamodels using the simulation results. Azaron et al. (2011) concerned with
the study of the constant due-date assignment policy in repetitive projects,
where the activity durations are exponentially distributed random variables and
they verified the results by Monte Carlo simulation. Akinnuli et al. (2012)
developed a computer-aided system which is based on a simulation and
empirical model for predicting job-shop FT and DD. Thurer et al. (2012)
evaluated the performance of DD setting rules in the context of complex
product structures by using simulation. In the following year Thurer et al.
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
5
(2013) compared the performances of 11 different DD Setting Rules in Job
Shops with Contingent Orders by the aid of simulation.
Hsieh et al. (2014) proposed a methodology, called progressive simulation
metamodeling (PSM). They used a response surface methodology based
simulation. Lee (2014) consider a problem of estimating order flow times in
two-stage hybrid flow shops, where orders arrive dynamically and various
scheduling schemes can be used. In this study several order flow time
estimation methods are devised and actual flow times of orders are obtained
from simulation runs. Li et al (2016) developed a simulation-based statistical
approach. They provided responsive and high-quality prediction of a new job's
FT through the system, which renders the capability of accurately quoting lead
times in real time. In this approach they integrated an analytical queuing
analysis, design of experiments, and statistical modeling to quantify the
dependence of a new job's FT distribution upon the shop status.
In the literature the common practice to predict the FT is using general job
and shop characteristics such as the number of the job in the queue, processing
times, number of resources, dispatching rules and etc. In addition, it is assumed
that the processing times or its probability distribution are known before
starting the production. However, this assumption is not valid for transformer
production. In this case study, the processing times or its probability
distribution is not known for the orders which will be produced the first time.
This causes problems at giving accurate price offers to the customers because
of unknown processing times and FT of unlimited kind of orders.
In the transformer production, the processing times of operations, the FT
and the DDs of each order vary according to the technical specifications that
are demanded by the customer. In this study, these parameters for each order is
predicted by using an interface software developed by the authors (which is
based on Arena simulation model, statistical modeling and probability
distributions) with the consideration of the technical data instead of using the
general job and shop characteristics. The following section gives a brief
description of the materials and methods used in this study. The case study,
simulation results and the interface software developed for the BEST
Transformers Co. are presented in the “Arena Simulation” section. Finally, the
conclusions are given in the last section.
Materials and Methods
The transformers production process has five main stages: core production,
winding operations, mechanical production, assembly, and final assembly. In
this study, simulation based interface software is developed for estimating the
processing times, FT and DD of the core production process of the
transformers to be produced at the first time. This interface software will be
used by the company to provide data to the sales staff at the bidding stage to
calculate the labor hours, labor cost and prepare better price offering. Also, the
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
6
outcomes of this software will be used for the scheduling of core production of
BEST Transformers Co.
To develop this interface software, the Arena simulation model is created
for the core production of the company. Then, the processing times of labor
intensive processes and preparation times are modeled by using probability
distributions while the standard machining processes are modeled by
mathematical models of design of experiment (DOE) techniques. To run this
simulation a user-friendly interface is coded by using Visual Studio software.
By using this interface, the user enters the technical design parameters to the
program and obtains the simulation results.
Arena Simulation
Simulation is a powerful tool to find good solutions in stochastic
environments. Rosetti (2010) defined the simulation as “a numerical technique
for conducting experiments on a digital computer which involves logical and
mathematical relationships that interact to describe the behavior of a system
over time”. Nowadays, simulation is often used in various fields and industries
by computational power and storage capacity of contemporary computers and
software which are better than ever.
There are many competing languages to conduct a simulation work in a
computer. Some of these languages are specially created for simulation and
uses drag and drop modules like Arena, Promodel, etc. Arena is one of the
most popular simulation software and has input and output analyzer modules.
Moreover, it exploits ActiveX Automation and Visual Basic for Applications
(VBA) for integrating directly with other programs. In this paper, Arena 14
simulation software is used for the simulation.
Response Surface Methodology (RSM)
The design of experiment (DOE) techniques are used for modeling the
relationship between the input variables (factors) and the output variable
(response) by using the minimum number of experimental results. By this way
it is possible to optimize the system parameters or to predict the response of
unpracticed combinations of different factor levels. This requires less time and
effort. The well-known DOE techniques are response surface methodology
(RSM), factorial design and Taguchi. RSM provides the mathematical relations
including interactions between the factors besides the linear and quadratic
relations. Equation (1) shows the general second-order polynomial RSM model
for the experimental design (Montgomery, 2001).
2
0
1 1
n n n
i i ii i ij i j
i i i j
Y X X X X
(1)
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
7
where Y is the response, Xi and Xj are the factors, terms, β0, βi, βii and βij are the
regression coefficients, and is the residual. Equation (1) may be written in
matrix notation as:
Y= β X +ε (2)
In this equation, Y and X represents the output and input matrices
respectively. The residual terms are given by the matrix represented with ε.
The least square estimator of the β matrix that composes of coefficients of the
regression equation (β0, βi, βii and βij) calculated by the given formula in
Equation (3), by ignoring ε;
1
T TX X X Y
(3)
where TX is the transpose of X.
Probability Distributions
The widely used probability distributions in the Arena are Uniform,
Normal, Exponential, Erlang, Gamma, Weibull, Lognormal, Beta, and
Triangular. In this study Erlang, Beta and Triangular distributions best fit the
processing times of labor intensive processes and preparation processes
(Rosetti, 2010). The common modeling situation for Triangular (a,m,b)
distribution assumes a minimum, a maximum and a most likely value. Also,
this distribution roughs model in the absence of data. The probability density
function (F(x)) of Triangular distribution is given in Equation (4-5):
2x a
b a b m
a x m (4)
2
1b x
b a b m
m x b (5)
Where a, b, and m are the minimum, maximum and mode respectively. The
common modeling situation for Erlang distribution is multiple phases of
service with each phase exponential. The probability density function of Erlang
( , r) distribution is given in Equation (6):
/1
0
/1
!
nxr
n
e x
n
(6)
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
8
Where is the mean of each of the component exponential distributions, and r
is the number of exponential random variables. Beta distribution is useful for
modeling task times in a bounded range with little data. The probability density
function of Beta ( 1 , 2 ) is given in Equation (7-8):
11 1
,
x x
0<x<1 (7)
Where is the complete Beta function given by:
1 11
0, = 1t t dt
(8)
Shape parameters Beta ( ) and Alpha ( ) specified as positive real numbers.
The Case Study
BEST Transformers is a transformer producer and produces power
transformers, distribution transformers and dry-type transformers. This study is
performed at a core production of distribution transformers. In this production
oil type and dry type cast resin transformers are produced by flexible flow shop
that is composed of consequent processes. The main production steps are core
production, winding operations, mechanical production, assembly, and final
assembly. The coils produced by separate winding operations are assembled
with the core. Then this mid-product is put in the tank which is produced by
mechanical production. Then, the final operations are performed on this mid-
product and the production is finalized. This study is motivated on core
production of oil type transformers.
In this production system, actually each of the received transformer order
can be accepted as a new product. The same transformers' designs with equal
power and voltage numbers may be completely different because of technical
specification those are demanded by the customers. These design differences
engage accurate prediction of processing times, FT and DD of the orders those
will be produced at first time.
In this study FT estimation of received transformer orders are performed
by using the mathematical equations of RSM, probability distributions and
Arena simulation. RSM was employed by using Minitab program, while
probability distributions and simulation were employed by using Arena
simulation software. In the proposed approach if the probability distribution for
processing time of the product has been defined, it is used at the simulation.
Otherwise, processing times of the products which are produced at the first
time are estimated by regression equations developed according to the product
technical specifications.
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
9
The core production is mainly composed of slicing, cutting and stacking
operations. In slicing operation, steel slice stocks are primarily checked and
ones which are not in stock are sliced in slice cutting machine from the roll in
width set design. In cutting operation, sliced rolls are cut in measurements
determined in design criteria and prepared for stacking operation. Stacking
operation is labor-intensive and the sheets cut are stacked in stacking tools by
workers. In addition, some parts used in stacking operations are produced in
carpenter. The results of the analyzes performed show that carpenter,
preparation and the final completion times don’t change slightly according to
technical characteristics, therefore probability distributions determined by the
Arena input analyzer software are used for these processing times. The
goodness of fit of the distributions is tested by the Chi-square test in 0.05
significance level, and the results show that determined probability
distributions can be used in the simulation. Cutting times are sensitive to
technical specifications according to the result of the analysis. Therefore, a
mathematical equation for cutting times is developed by RSM, and cutting
times generated from mathematical equations are used in the simulation. The
probability distributions and mathematical model used in the simulation are
given at Table 1 in terms of minutes. Transfer distances in the factory among
work stations and stock fields are measured and transfer times of material/
product are determined for the simulation by entering the speed of used
transporters.
Table 1. Processing Time Models in the Simulation
Operation Model
Carpenter Time 10+120*Beta (0.751, 1.31)
Tap Slice Time (20+119*Beta(1.44,1.5))
Cutting Time
(x1:weight, x2:number of sheet)
0.0495443-5.16233*x1+62.8709*x2-
143.455*x1*x1-81.0161*x2*x2 + 230.667*x1*x2
Preparation Time for Tap
Cutting
Triangular (8, 11, 14)
Completion Time for Cutting Triangular (23, 29, 35)
Preparation Time for Stacking Triangular (32, 49, 61)
Limb Stacking Time (6+Erlang(3.69, 2)
Completion Time for Stacking Triangular (30, 40, 50)
The simulation model firstly reads the production order records pending to
be produced from an excel file and simulates these orders by taking in to
account the dynamics conditions of the workshop. The orders whose required
processing time and DD need to be calculated are added current orders records
and simulated.
The created simulation model will be used by both the sales and
production departments. Specially, the sales staff is not familiar to simulation
studies. Therefore, a user-friendly interface is developed for the effective use
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
10
of the simulation in the firm. The developed interface software has two
modules: order analysis for sales department and scenario analysis for the
production department. The screenshot of the interface software is given at
Figure 1.
Figure 1. The Screenshot of the Interface Software
Sales staff enters technical specifications of current order to the simulation
model by the aid of interface software. Thus, processing times, FT and DD can
be estimated by the interface software embedded simulation model at a bidding
stage. On the other hand, Scenario analysis module is used by the production
department. The production department can make a scenario analysis for the
processing time, FT and DD by changing the existing order in simulation.
The 95% confidence intervals for mean of operation times in carpenter,
slicing, cutting and stacking operations, FT and DD for each group in the order
are calculated by simulation as the outputs of interface software. Moreover, the
best and the worst scenarios for FT and DD can be generated according to the
result of replications in simulation and relative probabilities related to the
completion time of each order group can be calculated by simulation as an
output of the software. An example of the output that the interface software
generates is given at Figure 2.
Figure 2. An Example Output of the Interface Software
An order list which includes 11 transformers in 6 groups is chosen for
the verification of outputs. Technical specifications of group orders are
completely different from each other and they cannot be given in this paper
because of the firm's privacy policy. The actual processing times of selected
groups of order are observed and given at Table 2. Simulation model runs
according to the actual order in production and the calculated 95% confidence
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
11
intervals for means of processing times are given at the same table in the terms
of minutes.
Table 2. Observed and Simulated Processing Times
Product
Code Frequency
Observed
Cabinetry
Time
95% Confidence
Interval
For Cabinetry
Time
Observed
Cutting Slice
Time
95% Confidence
Interval
For Cutting Slice
Time
Observed
Cutting
Time
95% Confidence
Interval
For Cutting Time
Observed
Stacking
Time
95% Confidence
Interval
For Stacking
Time
4907892 2 92 90.19 - 114.33 216 198.91 - 215.35 503 495.18 - 497. 61 1154 1152.3 - 1237.8
4907889 1 61 48.31 - 66.10 137 135.4 - 150.60 313 311.02 - 313.11 522 508.32 - 548.96
4907839 1 98 94.73 - 128.80 59 61.10 - 64.10 221 220.50 - 223.03 233 221.96 - 240.27
4906619 5 276 259.58 - 295.01 410 406.44 - 452.31 1153 1151.58 - 1156.94 1980 1960.5 - 2053.4
4906282 1 90 88.42 - 116.92 81 79.45 - 84.91 269 266.27 - 270.47 316 271.55 - 314.23
4907219 1 91 86.16 - 115.09 103 97.312 - 107.03 187 185.06 - 188.11 286 223.94 - 380.75
The observed FT and DD, and simulation results under 95 % confidence
level are calculated and given at Table 3.
Table 3. Observed and Simulated FT and DD
Product
Code Frequency
Observed
Flow Time
95% Confidence Interval
For FlowTime
Observed
Due Date
95% Confidence Interval
For Due Date
4907892 2 2156 2015 - 2199 2908 2900.3 - 2952.9
4907889 1 1285 1121 - 1376 1901 1800.7 - 1915.2
4907839 1 720 685 - 796 1978 1911.0 - 1991.0
4906619 5 4250 4086 - 4580 5320 5102.8 - 5765.6
4906282 1 987 885 - 1112 1820 1733.9 - 1834.1
4907219 1 825 751 - 916 2408 2359.3 - 2591.8
According to the Table 2 and 3, it is clearly indicated that the predicted
simulation results are between the bounds of the confidence interval.
Therefore, it is clear that the simulation model generates successful outputs.
The developed simulation model and interface can be used as decision
support software in the firm for both price offering at bidding stage and
production scheduling.
Conclusions
This study focused on predicting the processing times, FT and DD of the
customer orders those will be produced the first time in a transformer producer.
For this purpose Arena simulation is used. Previously presented studies on this
subject are used general job and shop characteristics and the processing times
are thought to be known before starting the production. The novelty proposed
by this study is using the technical specifications of the orders that are
demanded by the customer directly. In this study, for the labor-intensive
processes, probability distributions are used to predict the processing times
while the processing times of machining processes are predicted by using
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
12
regression equations. Also, interface software is developed. By using this
interface the user enters the technical design parameters to the program and
obtains the simulation results. The results indicate that using the technical
specifications directly gives good results and accurate prediction can be
performed.
References
Akinnuli, B. O., Oluwadare, S. A., and Aderoba, A. A. 2012. A computer-aided
system for job-shop flow time and due-date prediction. Advances in Materials
and Systems Technologies III Book Series: Advanced Materials Research. 367,
(Sep. 2012), 639-645.
Azaron, A., Fynes, B., and Modarres, M. 2011. Due date assignment in repetitive
projects. International Journal of Production Economics. 129, 1 (Jan. 2011), 79-
85. DOI= 10.1016/j.ijpe.2010.09.004.
Baykasoglu, A. and Gocken, M. 2009. Gene expression programming based due
date assignment in a simulated job shop. Expert Systems with Applications. 36, 10
(Dec. 2009), 12143-12150. DOI=10.1016/j.eswa.2009.03.061.
Ekren, B.Y. and Ornek, A.M. 2008. A simulation based experimental design to
analyze factors affecting production flow time. Simulation Modelling Practice
and Theory. 16, 3 (Mar. 2008), 278-293. DOI= 10.1016/j.simpat.2007.11.016.
Joseph, O.A. and Sridharan, R. 2011. Analysis of dynamic due-date assignment
models in a flexible manufacturing system. Journal of Manufacturing Systems.
30, 1 (Jan. 2011) 28-40. DOI= 10.1016/j.jmsy.2011.02.005.
Hsieh, L.Y., Chang, K.H., and Chien, C.F. 2014. Efficient development of cycle time
response surfaces using progressive simulation metamodeling. International
Journal of Production Research. 52, 10 (May 2014), 3097-3109. DOI= 10.1080/
00207543.2013.864055.
Lee, G.C. 2014. Real-time order flow time estimation methods for two-stage hybrid
flowshops, OMEGA-International Journal of Management Science. 42, 1 (Jan.
2014), 1-8. DOI= 10.1016/j.omega.2013.02.004.
Li, M., Yang, F., Wan, H., and et al. 2015. Simulation-based experimental design and
statistical modeling for lead time quotation. Journal of Manufacturing Systems.
37 (Oct. 2015), 362-374. DOI= 10.1016/j.jmsy.2014.07.012.
Montgomery, D.C., 2001. Design and Analysis of Experiments, 5th Edition, John
Wiley &Sons, Inc., New York, NY.
Rosetti, M.D. 2010. Simulation Modeling and Arena. 2nd ed., John Wiley&Sons, Inc.,
Hoboken, New Jersey, NJ.
Slomp, J., Bokhorst, J.A.C., and Caprihan, R. 2011. Average flow time estimation of
jobs in a flexible manufacturing cell consisting of a number of identical
machines. In proceedings of the IEEE International Conference on Industrial
Engineering and Engineering Management (IEEM) (Singapore, December 06-09,
2011). 1426-1430.
Thuerer, M., Huang, G., Stevenson, M., and et al. 2012. The performance of due
date setting rules in assembly and multi-stage job shops: an assessment
by simulation. International Journal of Production Research, 50, 20 (2012),
5949-5965. DOI= 10.1080/00207543.2011.638942.
Thuerer, M., Stevenson, M., Silva, C., and et al. 2013. Towards an integrated
workload control (WLC) concept: the performance of due date setting rules in job
ATINER CONFERENCE PAPER SERIES No: IND2016-1974
13
shops with contingent orders. International Journal of Production Research. 51,
15 (Aug. 2013), 4502-4516. DOI=10.1080/00207543.2013.774485.
Vinod, V. and Sridharan, R. 2011. Simulation modeling and analysis of due-
date assignment methods and scheduling decision rules in a dynamic job shop
production system. International Journal of Production Economics. 129, 1, (Jan.
2011), 127-146. DOI= 10.1016/j.ijpe.2010.08.017.